Project

Asset Degeneration Long Term

ProRail is responsible for the maintenance, extension, management, and safety of the national railway network in the Netherlands. Therefore, it is important for ProRail to get a good picture of the condition of the various assets, so that they can be better maintained and timely replaced. This helps them to keep the cost down while ensuring safety.

ProRail is responsible for the maintenance, extension, management, and safety of the national railway network in the Netherlands. Therefore, it is important for ProRail to get a good picture of the condition of the various assets, so that they can be better maintained and timely replaced. This helps them to keep the cost down while...
Project
Project Insight and Forecasting
Client
ProRail
Industry
Rail, Asset management
Project type
Prediction/Machine learning, Modelling, Data engineering/Cloud infrastructure
Geography
Netherlands
Year
2020-current
Website
prorail.nl

Challenge

ProRail is responsible for the maintenance, extension, management, and safety of the national railway network in the Netherlands. Therefore, it is important for ProRail to get a good picture of the condition of the various assets, so that they can be better maintained and timely replaced. This helps them to keep the cost down while ensuring safety. 

Replacements and maintenance are financed by the government. ProRail indicates 15 years in advance how many assets are expected to be replaced. If the lifespans of the assets are not accurately predicted, this can mean major financial differences versus the short-term planning that is made based on the actual real-world condition. The lifespans of the different assets were mainly based on expert estimates. Lynxx helps ProRail to improve these estimates by combining all available information about the specific assets in an advanced machine learning model. This limits the financial risk for ProRail and the government.

Lynxx has made algorithms for various asset types, but in this case we will focus on the rail tracks.

Our solution

The current condition of the rails is monitored by using measurements of the vertical and horizontal wear. To predict the future condition of the rails, these measurements are fed into a predictive model along with data about the type of material, number of trains passed etc. Because there is only limited condition data available for the ballast and sleeper, we assess the lifespan of these objects by their specifications and the expected amount of traffic. 

Lynxx has developed degeneration models for rails, ballast and sleepers that contribute to improved insights into service life. After a batch of new measurements comes in, the client can easily update the predictions by running the full-fledged pipeline in Azure.  These forecasts are used by rail experts to plan asset replacements for the next 5-7 years. In addition, they form the basis for ProRail’s long term (up to 15 years) grant applications to the government.

Results

The model currently makes predictions for 73% of the Dutch rail network. This is the part of the track that is being used the most, and is periodically measured. The accuracy of these predictions has been validated over a period of 8 years. The model shows a clear improvement compared to the previously used methods: it reduces the average error by more than 30%, while also being more flexible with fluctuations in traffic. The better insights into the wear of the track ensures that the replacements are more accurately planned.

Besides improving these predictions, we currently are also developing models which predict the likelihood of failures for several assets. The first application lies in improving the efficiency of the planned maintenance in switches. This helps ProRail to improve their scheduling, which will save cost and reduce the downtime of the entire rail network.

Travelers don’t have to use another platform or ticket machine to buy a ticket

Kim de Groot

Consultant
@ Lynxx
Nullam finibus in mauris eget malesuada. Pellentesque ipsum ante, elementum non dui sed, vehicula euismod ante. Proin efficitur diam dui, luctus congue mauris rutrum at. Etiam vel velit hendrerit, lobortis ligula vel, posuere magna. Sed aliquet convallis ipsum, nec molestie ante ultricies in. Aliquam eu odio egestas, pharetra ante at, pellentesque nulla. Fusce ac gravida ante, id volutpat neque. Nunc ut leo sed lectus ullamcorper convallis convallis vitae odio. Suspendisse potenti.
Travelers don’t have to use another platform or ticket machine to buy a ticket

Kim de Groot

Consultant
@ Lynxx
Nullam finibus in mauris eget malesuada. Pellentesque ipsum ante, elementum non dui sed, vehicula euismod ante. Proin efficitur diam dui, luctus congue mauris rutrum at. Etiam vel velit hendrerit, lobortis ligula vel, posuere magna. Sed aliquet convallis ipsum, nec molestie ante ultricies in. Aliquam eu odio egestas, pharetra ante at, pellentesque nulla. Fusce ac gravida ante, id volutpat neque. Nunc ut leo sed lectus ullamcorper convallis convallis vitae odio. Suspendisse potenti.